[
https://issues.apache.org/jira/browse/HBASE-21928?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16771881#comment-16771881
]
Duo Zhang commented on HBASE-21928:
-----------------------------------
OK, in sync client implementation, we will only set request to HIGH_QOS at
client, so it will not consume the meta handler. But anyway, the priority can
be set the user, we can not prevent user to set a very high priority and block
us...
> Introduce a MasterPriorityRpcScheduler and make it default
> ----------------------------------------------------------
>
> Key: HBASE-21928
> URL: https://issues.apache.org/jira/browse/HBASE-21928
> Project: HBase
> Issue Type: Bug
> Reporter: Zheng Hu
> Assignee: Duo Zhang
> Priority: Major
> Fix For: 3.0.0, 2.2.0, 2.3.0
>
> Attachments: HBASE-21928.patch
>
>
> [INFO] Running
> org.apache.hadoop.hbase.master.TestServerCrashProcedureCarryingMetaStuck
> [ERROR] Tests run: 1, Failures: 0, Errors: 1, Skipped: 0, Time elapsed:
> 41.604 s <<< FAILURE! - in
> org.apache.hadoop.hbase.master.TestServerCrashProcedureCarryingMetaStuck
> [ERROR]
> test(org.apache.hadoop.hbase.master.TestServerCrashProcedureCarryingMetaStuck)
> Time elapsed: 30.496 s <<< ERROR!
> java.util.concurrent.TimeoutException
> at
> org.apache.hadoop.hbase.master.TestServerCrashProcedureCarryingMetaStuck.test(TestServerCrashProcedureCarryingMetaStuck.java:93)
> As described in HBASE-21907, the implementation for HMaster is a bit
> complicated so we if mix the report region transition call and other priority
> calls in the same executor, it will be easy to lead to dead lock. So let's
> use MasterFifoRpcScheduler by default for HMaster.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
